Department-level change: Using social network analysis to map the hidden
structure of academic departments
Charles Henderson and Kathleen Quardokus
Citation: AIP Conf. Proc. 1513, 170 (2013); doi: 10.1063/1.4789679
View online: http://dx.doi.org/10.1063/1.4789679
View Table of Contents: http://proceedings.aip.org/dbt/dbt.jsp?KEY=APCPCS&Volume=1513&Issue=1
Published by the American Institute of Physics.
Additional information on AIP Conf. Proc.
Journal Homepage: http://proceedings.aip.org/
Journal Information: http://proceedings.aip.org/about/about_the_proceedings
Top downloads: http://proceedings.aip.org/dbt/most_downloaded.jsp?KEY=APCPCS
Information for Authors: http://proceedings.aip.org/authors/information_for_authors
Downloaded 29 Jan 2013 to 141.218.60.60. Redistribution subject to AIP license or copyright; see http://proceedings.aip.org/about/rights_permissions
Department-Level Change: Using Social Network Analysis
to Map the Hidden Structure of Academic Departments
Charles Henderson*† and Kathleen Quardokus*
*Mallinson Institute for Science Education, Western Michigan University, Kalamazoo, MI 49008, USA
†
Department of Physics, Western Michigan University, Kalamazoo, MI 49008, USA
Abstract. Efforts to improve teaching in higher education have often focused on individual faculty. However, there is a
growing consensus that the academic department is a more productive focus of change initiatives. Yet, academic
departments are not all the same. Understanding the structure of relationships within a department is important for
identifying who should be involved in the change effort and in what roles. It is also likely that a successful change effort
will modify the structure of relationships within a department. This paper presents the preliminary results from a study
of two academic departments at a research university. A social network for each department was constructed based on a
web survey that asked faculty to identify colleagues with whom they had teaching-related conversations. We identify
characteristics of the individuals and departments and describe how learning about this hidden structure can be
beneficial to change agents.
Keywords: Social Network Analysis, Educational Change, Higher Education.
PACS: 01.40.Fk
reform effort; however, not every faculty member
participated directly in the reform.
Data were collected using an online survey sent to
individual faculty members. The survey asked
respondents to identify others within the department
with whom they had discussions about teaching.
Respondent chose people from a dropdown list of
names (that included all faculty in the department) and
identified the frequency with which the teaching
discussions occurred over the last academic year
(never, less than once per month, monthly, weekly,
and nearly every day). The survey also included a freeresponse question to allow respondents to include
individuals outside of the department.
INTRODUCTION
In recent decades, educational researchers have
developed significant knowledge about instructional
practices that improve student learning in
undergraduate physics courses. Similar practices
appear to be effective across all science disciplines. In
efforts to promote the use of these improved
instructional practices, change agents based in the
sciences typically treat the problem by focusing on
individual faculty [1]. Yet, there is a growing body of
evidence that suggests the academic department is the
most productive unit of change [2].
Researchers, though, are only just beginning to
develop knowledge about how to work with the
department as a unit of change. This paper describes a
preliminary study of the use of social network analysis
as a tool to understand the hidden social structures of
academic departments. We first discuss the methods
used to model a departmental social network. We then
use examples of the social networks of two science
departments at a large research university to describe
how knowing about the network structure might be
useful for change agents.
RESULTS
Department 1 had a response rate of 44% and
Department 2 had a response rate of 52%. Responses
were analyzed using standard social network analysis
with the ORA software tool [3]. Links between faculty
members represent discussions about teaching. The
social networks were assumed to be undirected [4] and
non-respondents were treated the same as respondents
who reported no conversations about teaching [5].
The results will be presented in two parts. First, we
will discuss issues related to the community structure
of each department. Then, we will discuss the roles
that individuals play within the department. In both
cases, we will highlight aspects of the analysis that are
likely to be most useful for change agents.
METHODS
This preliminary study is based on the network
structure of teaching discussions in two science
departments at a research university. Both departments
were involved in a multi-departmental four-year
2012 Physics Education Research Conference
AIP Conf. Proc. 1513, 170-173 (2013); doi: 10.1063/1.4789679
© 2013 American Institute of Physics 978-0-7354-1134-0/$30.00
170
Downloaded 29 Jan 2013 to 141.218.60.60. Redistribution subject to AIP license or copyright; see http://proceedings.aip.org/about/rights_permissions
Department 2 (N=46): Although the survey was
only distributed to the 44 members of Department 2,
two non-department members were mentioned by
more than one respondent. These were added to the
network. The resulting network consisted of one main
component and 18 isolated faculty members. The
Newman
community
algorithm
found
four
communities in the main component (Figure 2). In
contrast to Department 1, Department 2 had a
modularity value of 0.35 which is close to the
threshold of values expected for networks that have
community structure. Therefore, the community
divisions in Department 2 were less prominent than in
Department 1. This lack of relatively isolated subgroups (with respect to discussions about teaching)
means that there are unlikely to be different ‘camps’ of
faculty within the department with disjointed views
about teaching and learning.
Part 1: Community Structure
The analysis of each department began with an
investigation of Newman Communities [6]. Newman
developed an algorithm for identifying partitions
(communities) within a network in such a way that the
number of links between individuals within a
community is as large as possible and the number of
links between individuals in different communities is
as small as possible. Communities within a larger
group are subdivisions of individuals who have more
conversations amongst themselves than with
individuals outside of the subdivision. Newman also
developed a measure of modularity to describe the
extent to which the network has modular/community
structure (on a scale of 0 to 1, with 0 representing no
community structure, or the structure that would be
expected in a randomly-created network with the same
number of individuals and links). According to
Newman, modularity values between .3 and .7 are
expected for networks that have community structure.
Networks with a modularity value greater than .7 are
rare and a modularity of less than .3 represents very
low community structure.
Department 1 (N=57): The resulting network
consisted of 19 isolated faculty members (no recorded
discussions about teaching) two pairs of dyads
(individuals that only discussed teaching with each
other) and the remaining 34 in a single component.
The Newman community algorithm identified four
subdivisions of the main component (Figure 1). The
modularity for Department 1 was .56, indicating a
significant level of community structure in the
department. The division of the network into
community structure has implications for identifying
key actors within the entire network and within
communities (discussed in the following section).
FIGURE 2.
Department 2’s main component and
community structure. Circles, diamonds, triangles and
squares represent the community membership.
Part 2: Important Individuals
Our main interest in understanding the structure of
discussions about teaching within these departments is
to inform educational change efforts. Understanding
this structure can allow change agents to identify
individuals who will be useful to spread information
throughout the network and to monitor the flow of
information across the network. Researchers have
identified three important roles that individuals can
play in a network (see Ref. [7] and Table 1): hubs,
pulse takers, and connectors. Individuals that play
these roles were identified using the ORA software
tool.
Hubs of information are individuals that have many
discussions about teaching. These members are central
FIGURE 1.
Department 1’s main component and
community structure. Circles, pentagons, triangles and
squares represent the community membership.
171
Downloaded 29 Jan 2013 to 141.218.60.60. Redistribution subject to AIP license or copyright; see http://proceedings.aip.org/about/rights_permissions
in the network. Hubs are important allies for reform
leaders because these individuals can quickly spread
information across the network. They can spread
information about procedures of the reform, as well as
reform successes (that may be important for recruiting
new science faculty into the reform).
Pulse takers have quick access to information in
the network. They have many links within a cluster,
but are not important for the flow of information
because they do not connect otherwise unconnected
individuals. Pulse takers do not spread information,
but are important for change agents since they can
provide the status of information and attitudes within
the network.
Connectors are individuals who act as gate keepers
between clusters (or hubs). These individuals can
choose to share information across the network, or
may restrict information flow by withholding
information. A change agent needs to have a
relationship with these individuals to ensure the flow
of information about the change effort is reaching
across the network. The change agent will also want
this information to represent the change effort in a
positive way.
Figures 3 and 4 show the top five hubs, pulse
takers, and connectors in each department. The
community structure must be taken into account when
identifying important individuals. For example, in
Department 1 the community structure was more
prominent. When identifying connectors, there were
no individuals recognized in the square community
(pictured in Figure 1). Because the community
structure is present in the network, it may be important
for the reform leader to include Faculty Member ‘CC’
(a connector in the square community and across
communities) in discussions about the flow of
information in the network. In departments that show
community structure, this extra analysis is necessary to
make sure all communities are represented in the
knowledge gained from and shared with the network.
FIGURE 3. Top 5 hubs, pulse takers, and connectors in
Department 1.
FIGURE 4. Top 5 hubs, pulse takers, and connectors in
Department 2.
TABLE 1. Three Important Roles of Individuals in Social Networks (adapted from Refs. [7,8]).
Role
Why Important?
How to Calculate [8]
Hubs
Have many discussions about teaching with
High total degree centrality (number of links divided by
many other individuals.
total number of possible links).
Pulse Takers
Have many discussions with well-connected
individuals, but are not themselves important
links in the flow of information.
High closeness centrality (the average of the shortest path
to all other network members) and low betweenness
centrality (how many shortest paths between pairs of
network members pass through the individual).
Connectors
Promote flow of information via discussions
with individuals in separated parts of the
network.
High betweenness centrality (how many shortest paths
between pairs of network members pass through the
individual).
172
Downloaded 29 Jan 2013 to 141.218.60.60. Redistribution subject to AIP license or copyright; see http://proceedings.aip.org/about/rights_permissions
CONCLUSIONS
DISCUSSION AND FUTURE WORK
Understanding the structure of the department’s
social network appears to be promising for informing
change efforts that focus on the department level as the
unit of change. We have shown how to identify
community structure and individuals within a network
(or community) that represent hubs, pulse takers and
connectors. These individuals can be targeted by
reform leaders to advance information throughout the
network and to be key actors in change efforts. We
have also presented preliminary evidence to suggest
that social network analysis may be a useful tool in
measuring alterations in departmental structure due to
change activities. This type of measurement is useful
for both documenting the impact of change initiatives
as well as improving theory related to creating change.
In this paper we have presented the preliminary
results of a study of the social network structure of
academic departments with respect to discussions
about teaching. Although we have discussed how to
measure community structure and individual roles, we
have not yet demonstrated how this knowledge is
useful for change agents. Here we present three
anecdotes to illustrate the potential utility of the type
of social network analysis presented here.
We have argued that information about the hidden
departmental structure can be important for change
agents. For this study, the information about these two
departments was given to the leader of the multidepartmental change effort. The leader indicated that
this information was very helpful in being able to
identify particular faculty members to include in the
change effort and for identifying the individuals that
were important to information flow.
In addition to being able to identify communities
within departments, a natural question is to ask why
such communities arise. Since these are departments at
a research university, one might think the communities
we identified for teaching-related discussions resulted
from the research subfields of these faculty. If this
were correct, then it would not be necessary to create a
separate network for teaching-related discussions. Yet,
there was only one community (out of the eight
communities we identified) that was strongly
associated with a research subfield. Through other
information we have about these departments, we
know that this research subfield has a weekly meeting
for the discussion of research issues, but at which
teaching discussions often occur before or after.
Finally, we suggest that this type of social network
analysis can be used to document the impact of change
efforts. For example, in Department 2, faculty
members were asked to recall the individuals with
whom they discussed teaching before the reform effort
began. The resulting network had a Newman
modularity value of .54, indicating that the department
previously had a much higher community structure
than it does currently (current modularity is .35). One
activity of the reform effort was the development of a
group of faculty who met biweekly throughout the
school year to discuss instructional concerns. The
impact of this meeting may have caused the decrease
in community structure of Department 2, into a more
centralized group. This reduction in community
structure is a positive sign for a change effort that
takes the department as the unit of change.
ACKNOWLEDGMENTS
We wish to thank the Howard Hughes Medical
Institute for support of this project.
REFERENCES
1. C. Henderson, A. Beach, and N. Finkelstein,
“Facilitating Change in Undergraduate STEM
Instructional Practices: An Analytic Review of the
Literature,” J. Res. Sci. Teach. 48(8), 952-984 (2011).
2. C. E. Wieman, K. K. Perkins, and S. Gilbert,
“Transforming Science Education at Large Research
Universities: A Case Study in Progress,” Change 42(2),
6-14 (2010).
3. http://www.casos.cs.cmu.edu/projects/ora/
4. This means that if one person reported having
discussions about teaching, then the return connection
was also created. If the strengths were not equal, the
highest frequency was assumed.
5. Both types of individuals could still have links to the
network, if they were identified by other respondents.
However, non-respondents add a limitation to the study
by increasing the possibility of unreported teaching
conversations between faculty members.
6. A. Clauset, M. Newman, and C. Moore, “Finding
community structure in very large networks,” Phys. Rev.
E, 70(6), 1-6 (2004).
7. K. Stephenson, “Trafficking in Trust: The Art and
Science of Human Knowledge Networks,” in
Enlightened Power: How Women are Transforming the
Practice of Leadership edited by L. Coughlin and E.
Wingard, San Francisco: Jossey-Bass. 2005, pp. 242265.
8. S. Wasserman and K. Faust, Social Network Analysis:
Methods and Applications, New York: Cambridge
University Press, 1994.
173
Downloaded 29 Jan 2013 to 141.218.60.60. Redistribution subject to AIP license or copyright; see http://proceedings.aip.org/about/rights_permissions